Ex Doc is correct.
This model uses the IMUC ITT survival curve as its basis. The only data points that we know for sure is the enrolment ramp and we know that exactly courtesy of NWBO and Dr. Bosch. Also we know that there were 100 patients alive when Dr. Bosch have his ASCO review.
This is the IMUC curve that I used to base my model. https://dl.dropboxusercontent.com/u/11047753/IMUC%20ITT%20GRAPH.JPG
With the enrolment ramp, each Patient was assigned an enrolment date.
A spreadsheet was developed with each row representing a patient randomized on the trial. There were 331 data rows. Patients where then assigned to treatment (experimental) vs Placebo arms in a ratio of 2 to 1 as per the protocol requires.
The event data from the IMUC control curve was expanded by ratios to represent the NWBO control MODEL. (since there were more patients on the NWBO trial than were on the IMUC trial). When one does this 87 events are required for the NWBO curve to be equal to the IMUC control curve. This is done manually by looking and measuring from the curve. The IMUC time data is in Days and that was converted to months for the NWBO model.
IMPORTANT: Please note that in the IMUC control curve there where no data points past 950 Days or 31.7 months. Therefore with this model nothing past 31.7 months would be accurate. In order to complete the
KM curve, one could not have any blanks so in my initial review for these not evented cases, I just subtracted the date of enrollment to the date of Bosch's 100 still alive talk. Since this NWBO trial has been going on for over a decade that gave some large follow-up numbers for some patients. But these did not affect this models outcome.
Below I adjusted this and the following link provides the updated curves for the NWBO survival model based upon IMUC dataset. Essentially nothing changes except there are no data points past 35 months because with this model we have no idea what is happening in that part of the curve.
Link for the new curve. https://dl.dropboxusercontent.com/u/11047753/Comparison%20Curve.JPG
The significant p value of .029 is still applicable.
Also with this model the Mean survival data is not a meaningful measure and should not be used in anyway since we do not know the amount of follow-up time for those cases that did not event yet.
Basically these curves show a reasonable model for the potential outcome of the trial up to 30 months time. After that this model is not able to predict the survival.
Basically, these curves shows that there are less events in the experimental arm than what would be predicted if the experimental had a survival outcome the same as the IMUC control curve.
Hope this helps you understand this model better.